Skip to content

Conversation

Zylphrex
Copy link
Member

The sum frame duration is not a useful metric without an idea of how frequently it's called. This swaps it out for avg frame duration. Would be better to have the number of calls as well but that's out of scope of this PR.

Closes PRO-25

The sum frame duration is not a useful metric without an idea of how frequently
it's called. This swaps it out for avg frame duration. Would be better to have
the number of calls as well but that's out of scope of this PR.
@Zylphrex Zylphrex requested a review from a team as a code owner September 15, 2025 22:46
Copy link

linear bot commented Sep 15, 2025

@github-actions github-actions bot added the Scope: Frontend Automatically applied to PRs that change frontend components label Sep 15, 2025
Copy link

codecov bot commented Sep 15, 2025

✅ All tests passed

cursor[bot]

This comment was marked as outdated.

@JonasBa
Copy link
Member

JonasBa commented Sep 16, 2025

The sum frame duration is not a useful metric without an idea of how frequently it's called.

I'd be curious to compare a few side by side and see which one surfaces the data more accurately. The reason I say this is because I am afraid that averages will hide outliers, and outliers are super important in the context of performance investigations.

Call frequency is not something we can accurately compute, but it would be great to expose like you said. I think we should let the user decide here, and make this configurable. Average seems like a sensible default and something that will work for most folks, but adding columns or changing what values they show could enable different types of investigations or views into the data

@Zylphrex
Copy link
Member Author

@JonasBa Yeah we don't have any accurate way of computing call frequency even on the backend today. The response currently contains the call frequency in the scope of the computed flamegraph but that's not super useful to expose for the similar reasons why sum of durations isn't useful to expose.

And while average does indeed hide high outliers, we don't actually have any way to identify them to begin with so that's a future problem like you said.

@Zylphrex Zylphrex enabled auto-merge (squash) September 16, 2025 15:08
@Zylphrex Zylphrex merged commit 1653532 into master Sep 16, 2025
45 checks passed
@Zylphrex Zylphrex deleted the txiao/feat/use-average-frame-duration-in-call-tree-table-view branch September 16, 2025 15:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Scope: Frontend Automatically applied to PRs that change frontend components
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants